thinQmaster Virtual testing partner Adaptive practice exams that find your weak spots and drill them. Every answer explained. From $69
Enroll

thinQmaster

Find your exam

Five questions free, before you decide.

Step 1 · choose a vendor

In one of our camps? Your practice exams come with the tuition — they are on your study page already.

AI & Machine Learning · Intermediate

FP32

32-bit single-precision floating point, the traditional default for neural network math.

Definition

FP32 offers high precision and range but uses 4 bytes per value. Modern training keeps master weights or accumulations in FP32 while computing in lower precision. TF32 is an NVIDIA Tensor Core format that uses FP32 range with reduced mantissa.

Exam tip

Bytes per parameter: FP32=4, FP16/BF16=2, FP8/INT8=1, FP4/INT4=0.5.

Related terms